ABSTRACT
The legal system has many legal texts related to different cases. It is challenging to manage and access every piece of data. Due to this, the need for text summarization is gaining more importance in research. This study presented a two-stage hybrid summarization model, wherein a term frequency-inverse document frequency–based extractive model was used to construct stage 1, followed by an encoder-decoder-based transformer model to generate a final abstractive summary for legal experts. This model was tested with 100 samples of civil documents from Indian Supreme Court data, and the effectiveness of the model was compared and analyzed using the Rouge score with other existing summarization models in the literature. The results show better system performance when compared with other existing summarization models.
